How Does the Mind Make Friend Choices?

by Jaimie Arona Krems & Daniel Conroy-Beam with thanks to Laureon A. Merrie


Lenù & Lila, Gene & Phineas, Thelma & Louise, Cap & Bucky. Best friends are fixtures in our lives. Best friends might not benefit our fitness as obviously and directly as mates do, but having good friends is thought to be the next best thing for one’s health behind quitting smoking. In a recent paper, we ask how the mind might integrate our myriad friend preferences—for friends who are smart like Lenù, charismatic like Phineas, steadfast like Thelma or Cap—to make actual friend choices.

Perhaps in an ideal world, everyone we liked would like us back, time and affection would be infinite, and everyone we’ve ever encountered (and liked) we could maintain as friends. But as nice as ideal worlds and adages about the unbounded nature of love can sound, they bump up against reality. For example, time is inelastic yet required to maintain social relationships. To some extent, our affections might also be finite. In the end, we can maintain only so many ties at any one time.

Somehow, then, we must evaluate, compare, and ultimately select friends. Considering how important having friends and their support can be for one’s survival—and potentially even the thriving of one’s offspring—these friend choices matter.

To make these choices, we need an algorithm that combines information about what we prefer in friends, and the extent to which each prospective friend fulfills those friend preferences, presumably translating this information into a summary rating of sorts. There are a number of “preference integration” algorithms that could accomplish this.

In the mating literature, a Euclidean integration hypothesis has performed quite well. Therein, the Euclidean algorithm represents mate preferences and prospective mates as points within an n-dimensional space, computing a summary mate value that is inversely proportional to the distance between those points. For example, this summary can integrate the discrepancies between wanting a friend who’s a 10/10 in loyalty, a 10/10 on intelligence, a 4/10 on optimism, and so on, with a prospective friend who is a 6/10 on loyalty and intelligence and an 8/10 on optimism.

We tested whether some of those critical predictions derived from a Euclidean integration hypothesis—which have already found support in the mating literature—hold in the friendship domain. To do this, we gathered data from three separate samples of US participants (N = 817 undergraduate and adult community participant convenience samples). In each sample, we asked participants to rate themselves on 23 trait characteristics using 8-point bipolar scales. For example, we’d ask people to rate themselves as being “Very loyal” to “Very disloyal”. Participants also completed ratings for both their ideal same-sex best friends and their actual same-sex best friends. They also repeated this for their ideal and actual same-sex close friends and for their ideal and actual (or most recent) romantic partners.

We then computed several dimensions of partner value, including the friend and mate values of each participant, as well as the extent to which each participant’s actual best friends (and close friends, mates) fulfilled their ideals. For example, a person’s best-friend preference fulfillment was calculated as the Euclidean distance between the preferences they marked having in a best friend and where they marked their actual best friends as falling on those trait characteristics.

Using these data, we found support for several critical predictions implied by a Euclidean integration hypothesis in the friendship domain. First, we predicted that people who possess characteristics that render them ideal friends (i.e., people with high friend value) should be highly sought after, meaning that they have their pick of friends. Thus, we predicted—and found—that high friend-value individuals seem able to attract friends who better fulfill their preferences. We also predicted and found that high friend-value individuals not only set higher standards for ideal friends—after all, they would seem able to make such demands—but also that they report having real-world best friends who themselves have higher friend-value. In addition, we also explored these patterns with respect to people’s mate value, mate preferences, and mate preference fulfillment—replicating Conroy-Beam and colleagues’ previous research.


“We also predicted and found that high friend-value individuals not only set higher standards for ideal friends, but also that they report having real-world best friends who themselves have higher friend-value”


We also started to explore something new: the dissociability of friend- and mate-value. On one hand, the characteristics that make one an ideal friend (e.g., being nice) can overlap with the characteristics that make one an ideal mate (e.g., being nice). At the same time, people might also have some different preferences for prospective friends and mates. A same-sex friend who is highly sexually attractive might be a threatening rival in the mating domain, for example, but an other-sex mate who is highly sexually attractive might be rather desirable.

So, do people’s friend values better predict their friend outcomes (than mate outcomes), and do people’s mate values better predict their mate outcomes (than friend outcomes)?

Sort of.

We find some, albeit mixed, support for dissociability. This finding is particularly intriguing, though, insofar as it adds to the broader conversation about association value (see Michael Bang Petersen, Aaron Sell, John Tooby, and Leda Cosmides’ 2012 paper in this same journal for more on that).

A large part of the impetus for our work was to address the gap in friendship research. To some extent, these important relationships remain an evolutionary mystery. More work on friendships could contribute to solving that puzzle. Luckily, a pretty handy playbook for conducting some future work on friendship already exists—the robust, deep literature on mating relationships.

Here, drawing from work conducted in the mate preferences literature, we found some support for the Euclidean integration hypothesis, suggesting that this algorithm might be how the mind integrates friend preferences to make friend choices. However, exciting new work recently accepted at EHB also suggests another algorithm to explore in friendships (see Bradner, Brase, & Huxman, in press). We look forward to pleasantly arguing with our friends about this, and to the future of work on friendship in evolutionary social science.



Read the paper: First tests of Euclidean preference integration in friendship: Euclidean friend value and power of choice on the friend market